The paper indicates that for TinyImageNet the accessible outliers are from ImageNet 22K with ImageNet 1K removed. Why would the Readme.md file in the ImageNet_dataset folder suggest the extraction of 800 classes complementary to the 200 classes included in the TinyImageNet from ImageNet 1K?????
The paper indicates that for TinyImageNet the accessible outliers are from ImageNet 22K with ImageNet 1K removed. Why would the Readme.md file in the ImageNet_dataset folder suggest the extraction of 800 classes complementary to the 200 classes included in the TinyImageNet from ImageNet 1K?????